Implicit conversion from float -4.875549048316245 to int loses precision